Which operator is better: "minus" or "not in"

Question

I have 2 queries below which produce same result, see Case1 and Case2.
My questions are:
1. When it is better to exclude records with MINUS operator, and when with "NOT IN" operator. Maybe we can theoretically answer that question? Maybe MINUS-operator will always be faster than "NOT IN"? Can you explain when to use which of those?
2. Why Case1 produces different row order than Case2?
Case1:
with T as
(select 1 ID, 'COMPLEX EXCLUDE CONDITIONS' CONDITION from dual union all
select 2 ID, 'INCLUDE' CONDITION from dual union all
select 3 ID, 'INCLUDE' CONDITION from dual)
select ID from T
minus
select ID from T
where CONDITION = 'COMPLEX EXCLUDE CONDITIONS'
2
3
*/Case2:
with T as
(select 1 ID, 'COMPLEX EXCLUDE CONDITIONS' CONDITION from dual union all
select 2 ID, 'INCLUDE' CONDITION from dual union all
select 3 ID, 'INCLUDE' CONDITION from dual)
select ID from T
where ID not in
   (select ID from T
   where CONDITION = 'COMPLEX EXCLUDE CONDITIONS')
3
2
*/

Answer 1

Hi,
MINUS does two full table scans & removes matches after whereas NOT IN does a full table scan of table 1 then for each row it searches through table two...assuming you have two proper tables e.g.:
TABLE1: 20,000 rows in 1000 blocks
TABLE2: 10,000 rows in 500 blocks
Reads required for minus:
Full scan of TABLE1 = 1000 blocks
+
Full scan of Table2 = 500 blocks
= 1500 reads
Reads required for NOT IN:
Full scan of TABLE1 = 1000 blocks
20,000 lookups in TABLE2 = 20,000 x (depth of index on TABLE2)
= 21,000 at least
So a lot more work is done with NOT IN. Taken from here. Note the gets:
SQL> select count(*) from
  2  ( select object_id from t1
  3    minus
  4    select object_id from t2
  5  )
  6  /
  COUNT(*)
       171
Statistics
          0  recursive calls
         24  db block gets
        136  consistent gets
         64  physical reads
          0  redo size
        380  bytes sent via SQL*Net to client
        518  bytes received via SQL*Net from client
          4  SQL*Net roundtrips to/from client
          3  sorts (memory)
          0  sorts (disk)
          1  rows processed
SQL> select count(*) from
  2  ( select object_id from t1
  3    where object_id not in
  4    ( select object_id from t2
  5    )
  6  )
  7  /
  COUNT(*)
       171
Statistics
          0  recursive calls
         12  db block gets
      84406  consistent gets
          0  physical reads
          0  redo size
        405  bytes sent via SQL*Net to client
        541  bytes received via SQL*Net from client
          4  SQL*Net roundtrips to/from client
          1  sorts (memory)
          0  sorts (disk)
          1  rows processedMike